JMIR Publications

ABSTRACT

Background:

Retrospective questionnaires are commonly used to assess exposure to unpleasant environmental stimuli such as sounds and odors but are vulnerable to recall and salience biases. Ecological momentary assessment (EMA) is often proposed as a superior alternative because it captures experiences close in time to their occurrence. However, EMA introduces feasibility and compliance challenges that may systematically distort exposure estimates.

Objective:

This study examined the feasibility, adoption, persistence, and completeness of smartphone-based EMA for documenting real-world exposure to unpleasant sounds and odors, with particular attention to compliance-related undercounting and bias in a population-based sample with a relatively older age distribution.

Methods:

EMA reporting was conducted over a three-month period alongside baseline and follow-up retrospective questionnaires. Participants with and without self-reported environmental sensitivity to noise or odors were offered access to a web-based EMA application for self-initiated reporting of exposure events. Feasibility outcomes included adoption, persistence, reporting volume, and self-reported missed events and barriers to use. Analyses were primarily descriptive.

Results:

Of 104 participants (environmentally sensitive [ES], n=42; nonenvironmentally sensitive [NES], n=62), EMA adoption was limited, with 15 ES participants (36%) and 22 NES participants (35%) submitting at least one report (χ²(1)=0.00, p=.98). Among EMA users, reporting persistence was short, with a mean reporting duration of 19.9 days (range 1–51) in the ES group and 7.9 days (range 1–32) in the NES group. Self‑reported underdocumentation was common: 14 of 15 ES users (93%) and 14 of 22 NES users (64%) reported experiencing one or more exposure events that were not recorded using EMA. In addition, many nonusers reported unrecorded exposure events during the study period (ES: 41%; NES: 13%). EMA reporting declined rapidly over time in both groups, with no participants maintaining reporting across the full three‑month period.

Conclusions:

EMA provided time-stamped and context-specific information unavailable from retrospective questionnaires but appeared to systematically undercount exposure experiences due to compliance-related limitations. These findings suggest that EMA shifts, rather than eliminates, measurement bias. We propose a conceptual model, the Compliance Filtering Model of EMA Bias, to describe how adoption, persistence, and reporting processes contribute to systematic underestimation of exposure. Compliance-related challenges may be particularly relevant in older populations or groups with lower familiarity with digital reporting tools. EMA should therefore be viewed as a complementary method whose limitations must be explicitly considered in study design and interpretation.